This study investigates the effects of aerosol-radiation interaction on subseasonal prediction using the Unified Forecast System (UFS) with an ocean, a sea ice and a wave component, coupled to an aerosol component. The aerosol component is from the current NOAA operational GEFSv12-Aerosols model, which includes the GOCART aerosol modules simulating sulfate, dust, black carbon, organic carbon, and sea-salt. The modeled aerosol optical depth (AOD) is compared to reanalysis from Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA2) and observations from Moderate Resolution Imaging Spectro-radiometer (MODIS) satellite andAtmospheric Tomography (ATom) aircraft. Despite biases primarily in dust and sea salt, a good agreement in AOD is achieved globally. The simulated radiative forcing (RF) from the total aerosols at the top of the atmosphere is approximately -2.5 W/m2 or -16 W/m2 per unit AOD globally. This is consistent with previous studies. In subsequent simulations, prognostic aerosol component is substituted with climatological aerosol concentrations derived from initial experiments. While regional differences in RF are noticeable in specific events between these two experiments, the resulting RF, surface temperature, geopotential height at 500 hPa and precipitation, show similarities in multi-year subseasonal applications. This suggests that given the current capacities of the aerosol modeling, adopting a climatology of aerosol concentrations as a cost-effective substitute for the intricate aerosol module may be a practical approach for subseasonal applications.

Wildland fires involve complicated processes that are challenging to represent in chemical transport models. Recent airborne measurements reveal remarkable chemical tomography in fresh wildland fire plumes, which remain yet to be fully explored using models. Here we present a high-resolution large eddy simulation (LES) model coupled to chemistry to study the chemical evolution in fresh wildland fire plume. The model is configured for a large fire heavily sampled during the Fire Influence on Regional to Global Environments and Air Quality (FIREX-AQ) field campaign, and a variety of airborne measurements are used to evaluate the chemical heterogeneity revealed by the model. We show that the model captures the observed cross-transect variations of a number of compounds quite well, including ozone (O3), nitrous acid (HONO), and peroxyacetyl nitrate (PAN), etc. The combined observational and modeling results suggest that the top and edges of fresh plume drive the photochemistry, while dark chemistry is also present but in the lower part of the plume. The model spatial resolution is shown to be very important as it may shift the chemical regime, leading to biases in O3 and NOx chemistry. Based on findings in this work, we speculate that the impact of small fires on air quality may be largely underestimated in models with coarse spatial resolutions.

Secondary organic aerosols (SOA) are formed from oxidation of hundreds of volatile organic compounds (VOCs) emitted from anthropogenic and natural sources. Accurate predictions of this chemistry are key for air quality and climate studies due to the large contribution of organic aerosols to submicron aerosol mass. Currently, only explicit models, such as the Generator for Explicit Chemistry and Kinetics of Organics in the Atmosphere (GECKO-A), can fully represent the chemical processing of thousands of organic species. However, their extreme computational cost prohibits their use in current chemistry-climate models, which rely on simplified empirical parameterizations to predict SOA concentrations. Recent applications of atmospheric chemistry emulation with machine learning (ML) applied to the simpler chemical mechanisms of tropospheric ozone have shown its ability to produce realistic predictions and significantly reduce the computational cost. This study proves that ML can accurately emulate SOA formation from an explicit chemistry model for several precursors with 100 to 100,000 times speedup over GECKO-A, making it computationally usable in a chemistry-climate model. To train the ML emulator, we generated thousands of GECKO-A box simulations sampled from a broad range of initial environmental conditions, and focused on the chemistry of three representative SOA precursors: the oxidation by OH of two anthropogenic (toluene, dodecane), and one biogenic VOC (alpha-pinene). We compare fully-connected and recurrent neural network methods and use an ensemble approach to quantify their underlying uncertainty and robustness. The SOA predictions generally remain stable over a simulation period of 5 days with an approximate error of 2-8\%.